Responsible AI in Gambling: Why the Industry Wants to Lead, Not Follow
G2E panel calls for industry-led AI standards
It is the opening day of the Global Gaming Expo in Las Vegas, and the research director of UNLV’s International Gaming Institute is making a claim that sounds almost old-fashioned in a room full of machine learning talk: the casino business should be the one setting the rules for artificial intelligence, not waiting to be handed them.
“The industry’s worked hard for decades to build up a reputation (and) gain trust,” said Kasra Ghaharian, pointing to gambling’s track record on anti-money-laundering and consumer-protection guidelines. “The same is true for now with AI. It’s a really good opportunity for the gaming industry to demonstrate, ‘This is how you do it.'”
That is the short version of why responsible AI in gambling is suddenly a live topic. The panel, titled “State of AI in Gaming 2026: Research Results on Industry Maturity, Regulatory Readiness, and the Road Ahead,” was moderated by Rick Arpin, KPMG’s U.S. gaming leader and Las Vegas managing partner, and featured Eric Bowers of Boyd Gaming, Simo Dragicevic of the IGI’s AI Research Hub, and Lori Kobashigawa of Fontainebleau Las Vegas.
The underlying research, produced by the IGI with KPMG LLP, found that most AI deployments in gaming sit in the back office and in security rather than in front of the customer, which puts the sector at a “developing” stage of adoption. It also flagged a gap that matters more than any product roadmap: regulators want to write rules for AI, but they do not yet have a clear picture of how operators are actually using it.
So no, the panel did not publish a numbered code of conduct. What it proposed was a posture. Be transparent with regulators, build on the compliance habits the industry already has, and close that knowledge gap before someone closes it for you. Everything below unpacks what that means for you as a player, starting with the misconceptions I hear most often.
What responsible AI means in gambling
Responsible AI in gambling means using automated systems in ways that are fair, explainable, auditable and protective of the player, with a human accountable for the outcome. In practice it covers four things: how algorithms treat players, what data gets collected, who can inspect the decisions, and what happens when the system gets it wrong.
Myth: it’s a marketing label with nothing behind it. Fair suspicion, and sometimes true. But the version being discussed at G2E has a concrete test attached to it, the same test that AML programmes face: can you show a regulator, on request, what the system does, what data it used, and who signed off? A principle you cannot evidence is a slogan. A principle with documentation, logging and an audit trail is a standard.
Algorithmic fairness in games
Myth: AI decides whether you win. This one needs killing off carefully, because it confuses two very different pieces of technology. Game outcomes in licensed casino games come from a certified random number generator, a piece of mathematics tested by independent labs and locked down by the licence conditions. It is deliberately not adaptive. It does not learn who you are, and it does not know your balance.
Algorithmic fairness in gaming is about the layer around the game: which promotion you get shown, what bonus terms you are offered, how a support dispute is triaged, whether a withdrawal gets flagged. Those are increasingly model-driven, and models can absolutely be unfair without anyone intending it. If a system is trained mostly on high-spending players, it will serve everyone else worse. That is the sort of bias responsible AI frameworks are designed to catch through testing and monitoring, not good intentions.
Data privacy and player rights
Every AI feature in gambling runs on your behaviour: session length, stake size, time of day, deposit patterns, which games you abandon. That data is genuinely useful for spotting harm early. It is also the most sensitive material an operator holds.
The rights worth knowing are simple. You should be able to find out what categories of data an operator collects and why, to withdraw consent for marketing uses, to ask for correction or deletion where the law allows it, and to get a human review of an automated decision that materially affects you, such as an account restriction. If an operator cannot explain its data practices in plain language, treat that as information about the operator.
Where AI already touches your play
AI is not arriving in gambling. It has been quietly working in the back rooms for years, which is exactly what the IGI and KPMG research describes. Here is the honest ledger of the main uses, the benefit to you, and what goes wrong without oversight.
| Application | What it does | Benefit to players | Risk without oversight |
|---|---|---|---|
| Fraud and AML monitoring | Flags suspicious payment and account patterns | Fewer stolen-card chargebacks, cleaner player pools | False positives freezing legitimate withdrawals |
| Harm detection models | Scores behaviour for signs of gambling harm | Earlier intervention, limit prompts, cool-off offers | Missed cases, or data used for marketing instead |
| Personalisation and hosting | Chooses offers, content and communication timing | Less irrelevant spam, smoother service | Targeting the most vulnerable with the most pressure |
| KYC and identity checks | Automates document and identity verification | Faster verification, quicker payouts | Biometric error rates hitting some groups harder |
| Game design analytics | Predicts which themes and mechanics land | Better games, less guesswork | Optimising purely for time on device |
The personalisation row is where the panel’s ambition and the player’s interest can pull in opposite directions. Bowers described a “dream scenario” of AI acting as an individual host for every player in a database, a “holistic personalized journey” across venue and online play. Dragicevic went further, imagining a “living database of players, simulating human behavior” within five to 10 years. Both are plausible. Both are also a description of extraordinarily precise influence over people’s gambling, which is precisely why the fairness and consent questions cannot be an afterthought.
Self-regulation versus waiting for the regulator
Myth: industry standards are just a way to dodge real rules. Sometimes they are. But the argument made at G2E is more practical than defensive. Regulators are going to write AI rules regardless. The only variable is whether those rules are written with a clear understanding of how the systems actually work, or without one.
Rules drafted in the dark tend to land in two bad ways. Either they are so vague that compliance means nothing, or they are so prescriptive that they outlaw useful things, like harm-detection models, along with harmful ones. Gambling has been through this cycle before with money laundering controls, and the industries that documented their own practices early ended up with workable requirements instead of blunt ones.
Self-regulation on its own is not enough, and nobody serious claims otherwise. The version that protects players is layered: published internal principles, independent testing, and a licensing regime with the power to audit and fine. Proactive standards do not replace external oversight. They make it competent.
Player protection through responsible AI
Myth: AI can reliably spot a problem gambler. It cannot, not with certainty, and any operator suggesting otherwise is overselling. These models are probabilistic. They produce risk scores from behavioural patterns, and they generate both false alarms and misses. A player on a tight budget can look identical to one in real trouble for weeks.
That limitation is the argument for doing it properly rather than not doing it. Responsible AI in player protection looks like this in practice:
- Escalation to humans. A risk score triggers a review or a conversation, never an automatic punishment.
- Marketing suppression that actually works. A player flagged as at risk, or self-excluded, is removed from promotional targeting across every channel, with no override for revenue reasons.
- Working tools, surfaced early. Deposit, loss and session limits, reality checks, cool-off and self-exclusion, offered at the moment behaviour changes rather than buried three menus deep.
- Explainability. If your account is limited or a payout is held, you get a reason and a route to appeal to a person.
- Separation of purpose. Data gathered to detect harm is not recycled to sell bonuses. This is the single clearest line between responsible and cynical use.
Note what none of this changes: the maths of the games. Responsible AI can make a platform safer and fairer to deal with. It does not alter the house edge. A slot at 96% RTP still carries a 4% edge over the long run, and every session remains a negative expectation bet. Player protection is about limiting damage, not about improving your odds.
What this means for players in India
India’s position is unusual because the regulatory conversation here has moved faster and harder than the AI conversation. The legal framework for real-money online gaming has tightened at the central level and continues to differ by state, so the first question for any Indian player is not “is this platform’s AI ethical” but “is this platform legal for me, and is it licensed anywhere that can hold it accountable”. Check the current position before you deposit rather than after.
Data rights, though, are now on firmer ground. Under India’s data protection law, personal data should be collected with informed consent for a stated purpose, kept only as long as needed, and subject to correction and erasure requests. That applies squarely to the behavioural data gambling platforms feed into their models. If an operator’s privacy notice is vague about profiling, automated decisions or third-party sharing, you are entitled to ask, and their answer tells you plenty.
Practical things worth doing: read the bonus terms and the wagering requirement before you accept an offer, because that is still where most player complaints start; set a deposit limit on day one rather than after a bad week; keep records of KYC submissions and withdrawal requests; and treat any platform that cannot give you a human being to appeal to as unsuitable, however sophisticated its technology looks.
If gambling has stopped feeling like entertainment, use the cool-off and self-exclusion tools on the platform and talk to someone. No algorithm is a substitute for that decision.
Frequently asked questions
What is responsible AI in gambling?
It is the practice of building and running automated systems in gambling so they are fair, transparent, auditable and accountable to a human. It covers algorithmic fairness in offers and decisions, data privacy, harm-detection models with human escalation, and the ability to explain any automated decision that affects a player’s account.
How does AI affect player protection?
Mostly through behavioural monitoring. Models scan play patterns for signs of harm and trigger limit prompts, cool-off offers or a staff review. They also power fraud detection and identity checks. The catch is that these models are probabilistic, so they need human oversight, appeal routes and strict rules against reusing harm data for marketing.
Why should the gaming industry lead on AI standards?
Because regulators will set rules either way, and the IGI and KPMG research found they currently lack visibility into how operators use AI. Standards written without that understanding risk being either toothless or so broad they block protective uses. Industry-led documentation and testing give regulators something workable to build on.
Does AI change whether I win?
No. Outcomes in licensed casino games come from certified random number generators that do not adapt to individual players. AI affects the surrounding experience, including offers, checks and support. The house edge stays exactly what the game’s RTP implies.
Tagged: G2E India industry standards player protection responsible AI responsible gambling
